DOCSLIB.ORG

Arxiv:1805.04488V5 [Math.NA]

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Arxiv:1805.04488V5 [Math.NA] GENERALIZED STANDARD TRIPLES FOR ALGEBRAIC LINEARIZATIONS OF MATRIX POLYNOMIALS∗ EUNICE Y. S. CHAN†, ROBERT M. CORLESS‡, AND LEILI RAFIEE SEVYERI§ Abstract. We define generalized standard triples X, Y , and L(z) = zC1 − C0, where L(z) is a linearization of a regular n×n −1 −1 matrix polynomial P (z) ∈ C [z], in order to use the representation X(zC1 − C0) Y = P (z) which holds except when z is an eigenvalue of P . This representation can be used in constructing so-called algebraic linearizations for matrix polynomials of the form H(z) = zA(z)B(z)+ C ∈ Cn×n[z] from generalized standard triples of A(z) and B(z). This can be done even if A(z) and B(z) are expressed in differing polynomial bases. Our main theorem is that X can be expressed using ℓ the coefficients of the expression 1 = Pk=0 ekφk(z) in terms of the relevant polynomial basis. For convenience, we tabulate generalized standard triples for orthogonal polynomial bases, the monomial basis, and Newton interpolational bases; for the Bernstein basis; for Lagrange interpolational bases; and for Hermite interpolational bases. We account for the possibility of common similarity transformations. Key words. Standard triple, regular matrix polynomial, polynomial bases, companion matrix, colleague matrix, comrade matrix, algebraic linearization, linearization of matrix polynomials. AMS subject classifications. 65F15, 15A22, 65D05 1. Introduction. A matrix polynomial P (z) ∈ Fm×n[z] is a polynomial in the variable z with coef- ficients that are m by n matrices with entries from the field F. We will use F = C, the field of complex k numbers, in this paper. Typically an expression in the monomial basis φk(z) = z is given for P (z), and often only regular matrix polynomials are considered, that is, with m = n (we will use n for the dimension) and where det P (z) is not identically zero. Matrix polynomials have many applications and their study is of both classic and ongoing interest. See the classic work [20] and the surveys [21] and [29] for theory and applications. In this paper, as is done in [2], we consider the case when any polynomial basis φk(z) is used. We require that the set {φk(z)} for 0 ≤ k ≤ ℓ forms a basis for polynomials of grade ℓ. The word “grade” is short for “degree at most.” Thus, we write our regular matrix polynomial as ℓ (1.1) P (z)= Pkφk(z) , k X=0 n×n 1 where the matrices Pk ∈ F are square, and the degree of P (z) is at most ℓ (P (z) has grade ℓ). arXiv:1805.04488v5 [math.NA] 25 Mar 2021 ∗Received by the editors on Month/Day/Year. Accepted for publication on Month/Day/Year. Handling Editor: Name of Handling Editor. Corresponding Author: Name of Corresponding Author †Centre for Medical Evidence, Decision Integrity and Clinical Impact (MEDICI Centre), Department of Anesthesia and Perioperative Medicine, Schulich School of Medicine and Dentistry, Western University ([email protected]) ‡Ontario Research Centre for Computer Algebra, School of Mathematical and Statistical Sciences, Western University ([email protected]) §Ontario Research Centre for Computer Algebra, School of Mathematical and Statistical Sciences, Department of Applied Mathematics, Western University (lrafi[email protected]) 1The degree of a matrix polynomial P (z) is defined as follows. If P (z) is identically zero, the degree is −∞. Otherwise, if P ℓ A k A P (z)= Pk=0 kz expressed in the monomial basis and ℓ is not the zero matrix, then the degree of (z) is ℓ. 1 The notion of “grade” is useful even for the monomial basis, but it is especially useful if the basis is n n k n−k an interpolational basis or the Bernstein basis φk(z) = Bk (z) = k z (1 − z) , when the degree of the polynomial may not be clear from the data. For more information about matrix polynomials, consult [27]. See also [29], [21], and consult the seminal book [20]. Linearizations using different polynomial bases were first systematically studied in [2]; for a more up-to-date treatment see [5]. Some other recent papers of interest include [4], [28], [12], [17], and [32]; this is a very active area. See also [1]. In that paper, standard triples for structured matrices are studied. In [9] several proofs are given of strict equivalence of various linearizations to the standard second linearization for the monomial basis, which we will use without much further comment in this paper. 1.1. Organization of the Paper. In Section 1.2, we establish notation, give the definitions of algebraic linearization and of generalized standard triples. We define this last in Definition 1.1 with reference to the representation in Equation (1.7). In Section 1.3, we show how to use a generalized standard triple in the construction of algebraic linearizations. We also give a proof, by construction of the necessary equivalence matrices E(z) and F (z), that algebraic linearizations are local linearizations (a very useful notion from [13]) in the same sense that its components are. In particular, if the components A and B are linearizations in the sets ΣA and ΣB respectively, then the “algebraic linearization” is a linearization in the intersection ΣA ∩ ΣB. In Section 2, we prove our main result, giving a universal expression for generalized standard triples for the linearizations using the polynomial bases that appear in Sections 3.1, 3.2, and 3.3. In Section 3, we introduce all the polynomial bases by scalar examples. We sum up in the final section. 1.2. Notation and Definitions. Two matrix polynomials P1(z) and P2(z) are called unimodularly equivalent if there exist unimodular matrix polynomials (that is, matrix polynomials with constant nonzero determinant) E(z) and F (z) with P1(z) = E(z)P2(z)F (z). A matrix pencil L(z) := zC1 − C0 is called a linearization of the matrix polynomial P (z) if both C1 and C0 are of dimension N ≥ n and L(z) is unimodularly equivalent to the block diagonal matrix diag(P (z), IN−n). Two linearizations Lm(z) and Lφ(z) are called strictly equivalent if the corresponding matrices are equivalent in the following stronger sense: C1,m = EC1,φF and C0,m = E(z)C0,φF (z), with the same invertible (constant) matrices E and F . In the paper [13] we find the powerful notion of a local linearization on a subset Σ ⊂ F. In the notation of their Definition 2.1, two rational matrices G1(z) and G2(z) are said to be equivalent in a nonempty set Σ if there exist matrices R1(z) and R2(z) each nonsingular in Σ with R1(z)G1(z)R2(z) = G2(z). This notion, intended for use with matrix rational functions, is also useful for matrix polynomials, as we will see. There is a related idea to linearization, that of “companion pencil” (A, B), where the only requirement is that det P (z) = K det (zB − A) for some nonzero constant K. A companion pencil, therefore, has the same eigenvalues as the matrix polynomial. Companion pencils that are not linearizations do not necessarily preserve eigenvectors or elementary divisors, and are less useful than linearizations. The usual reversal2 of a matrix polynomial of grade ℓ is the polynomial rev P (z) = zℓP (z−1). A 2This definition, which is standard, is particularly appropriate for the monomial basis. The coefficients of the reversed matrix polynomial in the monomial basis are simply the same matrices in reverse order. The notion of a reversal, however, is independent of the basis used, and indeed reversals can be done differently. In [10] for instance we find a slightly different definition of reversal, appropriate for computation in a Lagrange or Hermite interpolational basis, which maps an arbitrary finite point to infinity; this difference allows for greater numerical stability. 2 linearization L(z) = zC1 − C0 of P is called a strong linearization if rev L(z) = C1 − zC0 is also a linearization of rev P (z). N×N If L(z) = zC1 − C0 ∈ C [z] is a linearization of P (z) in z ∈ Σ, then as a necessary consequence det(P (z)) = q(z)det(L(z)) = det(zC1 − C0) for some rational q(z) whose zeros and poles lie outside Σ. The eigenvalues of P in Σ are thus computable from the generalized eigenvalues of L. For this linearized problem several standard methods are available. A standard pair (X, T ) for a regular matrix polynomial P (z) expressed in the monomial basis with coefficients Pk is defined in [20] or in [27] as having the following properties: X has dimension n × nℓ, T has dimension nℓ × nℓ, ℓ k (1.2) PkXT =0 , k X=0 and that the nℓ by nℓ matrix X XT (1.3) Q = . . XT ℓ−1 is nonsingular. We can then define a third matrix 0n . (1.4) Y = Q−1 . 0 n I n and say that the triple (X, T , Y ) is a standard triple for a monic P (z). It is pointed out in [27] that monicity of P (z) is not required for many of the formulæ to do with standard pairs (but is required for some). Theorem 12.1.4 of [19] states that if there are matrices X, T , and Y of dimension n × nℓ, nℓ × nℓ, and nℓ × n for which −1 −1 (1.5) P (z)= X(zIn − T ) Y then (X, T , Y ) is a standard triple for P (z). This is also reported in Lemma 2 in [1]. There are two other representations of a matrix polynomial given a standard triple: the right canonical form, and the left canonical form.
Load more

Recommended publications
  • Matrix Analysis (Lecture 1)
    Matrix Analysis (Lecture 1) Yikun Zhang∗ March 29, 2018 Abstract Linear algebra and matrix analysis have long played a fundamen- tal and indispensable role in fields of science. Many modern research projects in applied science rely heavily on the theory of matrix. Mean- while, matrix analysis and its ramifications are active fields for re- search in their own right. In this course, we aim to study some fun- damental topics in matrix theory, such as eigen-pairs and equivalence relations of matrices, scrutinize the proofs of essential results, and dabble in some up-to-date applications of matrix theory. Our lecture will maintain a cohesive organization with the main stream of Horn's book[1] and complement some necessary background knowledge omit- ted by the textbook sporadically. 1 Introduction We begin our lectures with Chapter 1 of Horn's book[1], which focuses on eigenvalues, eigenvectors, and similarity of matrices. In this lecture, we re- view the concepts of eigenvalues and eigenvectors with which we are familiar in linear algebra, and investigate their connections with coefficients of the characteristic polynomial. Here we first outline the main concepts in Chap- ter 1 (Eigenvalues, Eigenvectors, and Similarity). ∗School of Mathematics, Sun Yat-sen University 1 Matrix Analysis and its Applications, Spring 2018 (L1) Yikun Zhang 1.1 Change of basis and similarity (Page 39-40, 43) Let V be an n-dimensional vector space over the field F, which can be R, C, or even Z(p) (the integers modulo a specified prime number p), and let the list B1 = fv1; v2; :::; vng be a basis for V.
    [Show full text]
  • 10 7 General Vector Spaces Proposition & Definition 7.14
    10 7 General vector spaces Proposition & Definition 7.14. Monomial basis of n(R) P Letn N 0. The particular polynomialsm 0,m 1,...,m n n(R) defined by ∈ ∈P n 1 n m0(x) = 1,m 1(x) = x, . ,m n 1(x) =x − ,m n(x) =x for allx R − ∈ are called monomials. The family =(m 0,m 1,...,m n) forms a basis of n(R) B P and is called the monomial basis. Hence dim( n(R)) =n+1. P Corollary 7.15. The method of equating the coefficients Letp andq be two real polynomials with degreen N, which means ∈ n n p(x) =a nx +...+a 1x+a 0 andq(x) =b nx +...+b 1x+b 0 for some coefficientsa n, . , a1, a0, bn, . , b1, b0 R. ∈ If we have the equalityp=q, which means n n anx +...+a 1x+a 0 =b nx +...+b 1x+b 0, (7.3) for allx R, then we can concludea n =b n, . , a1 =b 1 anda 0 =b 0. ∈ VL18 ↓ Remark: Since dim( n(R)) =n+1 and we have the inclusions P 0(R) 1(R) 2(R) (R) (R), P ⊂P ⊂P ⊂···⊂P ⊂F we conclude that dim( (R)) and dim( (R)) cannot befinite natural numbers. Sym- P F bolically, we write dim( (R)) = in such a case. P ∞ 7.4 Coordinates with respect to a basis 11 7.4 Coordinates with respect to a basis 7.4.1 Basis implies coordinates Again, we deal with the caseF=R andF=C simultaneously. Therefore, letV be an F-vector space with the two operations+ and .
    [Show full text]
  • Selecting a Monomial Basis for Sums of Squares Programming Over a Quotient Ring
    Selecting a Monomial Basis for Sums of Squares Programming over a Quotient Ring Frank Permenter1 and Pablo A. Parrilo2 Abstract— In this paper we describe a method for choosing for λi(x) and s(x), this feasibility problem is equivalent to a “good” monomial basis for a sums of squares (SOS) program an SDP. This follows because the sum of squares constraint formulated over a quotient ring. It is known that the monomial is equivalent to a semidefinite constraint on the coefficients basis need only include standard monomials with respect to a Groebner basis. We show that in many cases it is possible of s(x), and the equality constraint is equivalent to linear to use a reduced subset of standard monomials by combining equations on the coefficients of s(x) and λi(x). Groebner basis techniques with the well-known Newton poly- Naturally, the complexity of this sums of squares program tope reduction. This reduced subset of standard monomials grows with the complexity of the underlying variety, as yields a smaller semidefinite program for obtaining a certificate does the size of the corresponding SDP. It is therefore of non-negativity of a polynomial on an algebraic variety. natural to explore how algebraic structure can be exploited I. INTRODUCTION to simplify this sums of squares program. Consider the Many practical engineering problems require demonstrat- following reformulation [10], which is feasible if and only ing non-negativity of a multivariate polynomial over an if (2) is feasible: algebraic variety, i.e. over the solution set of polynomial Find s(x) equations.
    [Show full text]
  • A Singularly Valuable Decomposition: the SVD of a Matrix Dan Kalman
    A Singularly Valuable Decomposition: The SVD of a Matrix Dan Kalman Dan Kalman is an assistant professor at American University in Washington, DC. From 1985 to 1993 he worked as an applied mathematician in the aerospace industry. It was during that period that he first learned about the SVD and its applications. He is very happy to be teaching again and is avidly following all the discussions and presentations about the teaching of linear algebra. Every teacher of linear algebra should be familiar with the matrix singular value deco~??positiolz(or SVD). It has interesting and attractive algebraic properties, and conveys important geometrical and theoretical insights about linear transformations. The close connection between the SVD and the well-known theo1-j~of diagonalization for sylnmetric matrices makes the topic immediately accessible to linear algebra teachers and, indeed, a natural extension of what these teachers already know. At the same time, the SVD has fundamental importance in several different applications of linear algebra. Gilbert Strang was aware of these facts when he introduced the SVD in his now classical text [22, p. 1421, obselving that "it is not nearly as famous as it should be." Golub and Van Loan ascribe a central significance to the SVD in their defini- tive explication of numerical matrix methods [8, p, xivl, stating that "perhaps the most recurring theme in the book is the practical and theoretical value" of the SVD. Additional evidence of the SVD's significance is its central role in a number of re- cent papers in :Matlgenzatics ivlagazine and the Atnericalz Mathematical ilironthly; for example, [2, 3, 17, 231.
    [Show full text]
  • QUADRATIC FORMS and DEFINITE MATRICES 1.1. Definition of A
    QUADRATIC FORMS AND DEFINITE MATRICES 1. DEFINITION AND CLASSIFICATION OF QUADRATIC FORMS 1.1. Definition of a quadratic form. Let A denote an n x n symmetric matrix with real entries and let x denote an n x 1 column vector. Then Q = x’Ax is said to be a quadratic form. Note that a11 ··· a1n . x1 Q = x´Ax =(x1...xn) . xn an1 ··· ann P a1ixi . =(x1,x2, ··· ,xn) . P anixi 2 (1) = a11x1 + a12x1x2 + ... + a1nx1xn 2 + a21x2x1 + a22x2 + ... + a2nx2xn + ... + ... + ... 2 + an1xnx1 + an2xnx2 + ... + annxn = Pi ≤ j aij xi xj For example, consider the matrix 12 A = 21 and the vector x. Q is given by 0 12x1 Q = x Ax =[x1 x2] 21 x2 x1 =[x1 +2x2 2 x1 + x2 ] x2 2 2 = x1 +2x1 x2 +2x1 x2 + x2 2 2 = x1 +4x1 x2 + x2 1.2. Classification of the quadratic form Q = x0Ax: A quadratic form is said to be: a: negative definite: Q<0 when x =06 b: negative semidefinite: Q ≤ 0 for all x and Q =0for some x =06 c: positive definite: Q>0 when x =06 d: positive semidefinite: Q ≥ 0 for all x and Q = 0 for some x =06 e: indefinite: Q>0 for some x and Q<0 for some other x Date: September 14, 2004. 1 2 QUADRATIC FORMS AND DEFINITE MATRICES Consider as an example the 3x3 diagonal matrix D below and a general 3 element vector x. 100 D = 020 004 The general quadratic form is given by 100 x1 0 Q = x Ax =[x1 x2 x3] 020 x2 004 x3 x1 =[x 2 x 4 x ] x2 1 2 3 x3 2 2 2 = x1 +2x2 +4x3 Note that for any real vector x =06 , that Q will be positive, because the square of any number is positive, the coefficients of the squared terms are positive and the sum of positive numbers is always positive.
    [Show full text]
  • 1 Expressing Vectors in Coordinates
    Math 416 - Abstract Linear Algebra Fall 2011, section E1 Working in coordinates In these notes, we explain the idea of working \in coordinates" or coordinate-free, and how the two are related. 1 Expressing vectors in coordinates Let V be an n-dimensional vector space. Recall that a choice of basis fv1; : : : ; vng of V is n the same data as an isomorphism ': V ' R , which sends the basis fv1; : : : ; vng of V to the n standard basis fe1; : : : ; eng of R . In other words, we have ' n ': V −! R vi 7! ei 2 3 c1 6 . 7 v = c1v1 + ::: + cnvn 7! 4 . 5 : cn This allows us to manipulate abstract vectors v = c v + ::: + c v simply as lists of numbers, 2 3 1 1 n n c1 6 . 7 n the coordinate vectors 4 . 5 2 R with respect to the basis fv1; : : : ; vng. Note that the cn coordinates of v 2 V depend on the choice of basis. 2 3 c1 Notation: Write [v] := 6 . 7 2 n for the coordinates of v 2 V with respect to the basis fvig 4 . 5 R cn fv1; : : : ; vng. For shorthand notation, let us name the basis A := fv1; : : : ; vng and then write [v]A for the coordinates of v with respect to the basis A. 2 2 Example: Using the monomial basis f1; x; x g of P2 = fa0 + a1x + a2x j ai 2 Rg, we obtain an isomorphism ' 3 ': P2 −! R 2 3 a0 2 a0 + a1x + a2x 7! 4a15 : a2 2 3 a0 2 In the notation above, we have [a0 + a1x + a2x ]fxig = 4a15.
    [Show full text]
  • Tight Monomials and the Monomial Basis Property
    PDF Manuscript TIGHT MONOMIALS AND MONOMIAL BASIS PROPERTY BANGMING DENG AND JIE DU Abstract. We generalize a criterion for tight monomials of quantum enveloping algebras associated with symmetric generalized Cartan matrices and a monomial basis property of those associated with symmetric (classical) Cartan matrices to their respective sym- metrizable case. We then link the two by establishing that a tight monomial is necessarily a monomial defined by a weakly distinguished word. As an application, we develop an algorithm to compute all tight monomials in the rank 2 Dynkin case. The existence of Hall polynomials for Dynkin or cyclic quivers not only gives rise to a simple realization of the ±-part of the corresponding quantum enveloping algebras, but also results in interesting applications. For example, by specializing q to 0, degenerate quantum enveloping algebras have been investigated in the context of generic extensions ([20], [8]), while through a certain non-triviality property of Hall polynomials, the authors [4, 5] have established a monomial basis property for quantum enveloping algebras associated with Dynkin and cyclic quivers. The monomial basis property describes a systematic construction of many monomial/integral monomial bases some of which have already been studied in the context of elementary algebraic constructions of canonical bases; see, e.g., [15, 27, 21, 5] in the simply-laced Dynkin case and [3, 18], [9, Ch. 11] in general. This property in the cyclic quiver case has also been used in [10] to obtain an elementary construction of PBW-type bases, and hence, of canonical bases for quantum affine sln. In this paper, we will complete this program by proving this property for all finite types.
    [Show full text]
  • Polynomial Interpolation CPSC 303: Numerical Approximation and Discretization
    Lecture Notes 2: Polynomial Interpolation CPSC 303: Numerical Approximation and Discretization Ian M. Mitchell [email protected] http://www.cs.ubc.ca/~mitchell University of British Columbia Department of Computer Science Winter Term Two 2012{2013 Copyright 2012{2013 by Ian M. Mitchell This work is made available under the terms of the Creative Commons Attribution 2.5 Canada license http://creativecommons.org/licenses/by/2.5/ca/ Outline • Background • Problem statement and motivation • Formulation: The linear system and its conditioning • Polynomial bases • Monomial • Lagrange • Newton • Uniqueness of polynomial interpolant • Divided differences • Divided difference tables and the Newton basis interpolant • Divided difference connection to derivatives • Osculating interpolation: interpolating derivatives • Error analysis for polynomial interpolation • Reducing the error using the Chebyshev points as abscissae CPSC 303 Notes 2 Ian M. Mitchell | UBC Computer Science 2/ 53 Interpolation Motivation n We are given a collection of data samples f(xi; yi)gi=0 n • The fxigi=0 are called the abscissae (singular: abscissa), n the fyigi=0 are called the data values • Want to find a function p(x) which can be used to estimate y(x) for x 6= xi • Why? We often get discrete data from sensors or computation, but we want information as if the function were not discretely sampled • If possible, p(x) should be inexpensive to evaluate for a given x CPSC 303 Notes 2 Ian M. Mitchell | UBC Computer Science 3/ 53 Interpolation Formulation There are lots of ways to define a function p(x) to approximate n f(xi; yi)gi=0 • Interpolation means p(xi) = yi (and we will only evaluate p(x) for mini xi ≤ x ≤ maxi xi) • Most interpolants (and even general data fitting) is done with a linear combination of (usually nonlinear) basis functions fφj(x)g n X p(x) = pn(x) = cjφj(x) j=0 where cj are the interpolation coefficients or interpolation weights CPSC 303 Notes 2 Ian M.
    [Show full text]
  • Math 623: Matrix Analysis Final Exam Preparation
    Mike O'Sullivan Department of Mathematics San Diego State University Spring 2013 Math 623: Matrix Analysis Final Exam Preparation The final exam has two parts, which I intend to each take one hour. Part one is on the material covered since the last exam: Determinants; normal, Hermitian and positive definite matrices; positive matrices and Perron's theorem. The problems will all be very similar to the ones in my notes, or in the accompanying list. For part two, you will write an essay on what I see as the fundamental theme of the course, the four equivalence relations on matrices: matrix equivalence, similarity, unitary equivalence and unitary similarity (See p. 41 in Horn 2nd Ed. He calls matrix equivalence simply \equivalence."). Imagine you are writing for a fellow master's student. Your goal is to explain to them the key ideas. Matrix equivalence: (1) Define it. (2) Explain the relationship with abstract linear transformations and change of basis. (3) We found a simple set of representatives for the equivalence classes: identify them and sketch the proof. Similarity: (1) Define it. (2) Explain the relationship with abstract linear transformations and change of basis. (3) We found a set of representatives for the equivalence classes using Jordan matrices. State the Jordan theorem. (4) The proof of the Jordan theorem can be broken into two parts: (1) writing the ambient space as a direct sum of generalized eigenspaces, (2) classifying nilpotent matrices. Sketch the proof of each. Explain the role of invariant spaces. Unitary equivalence (1) Define it. (2) Explain the relationship with abstract inner product spaces and change of basis.
    [Show full text]
  • Arxiv:1901.01378V2 [Math-Ph] 8 Apr 2020 Where A(P, Q) Is the Arithmetic Mean of the Vectors P and Q, G(P, Q) Is Their P Geometric Mean, and Tr X Stands for Xi
    MATRIX VERSIONS OF THE HELLINGER DISTANCE RAJENDRA BHATIA, STEPHANE GAUBERT, AND TANVI JAIN Abstract. On the space of positive definite matrices we consider dis- tance functions of the form d(A; B) = [trA(A; B) − trG(A; B)]1=2 ; where A(A; B) is the arithmetic mean and G(A; B) is one of the different versions of the geometric mean. When G(A; B) = A1=2B1=2 this distance is kA1=2− 1=2 1=2 1=2 1=2 B k2; and when G(A; B) = (A BA ) it is the Bures-Wasserstein metric. We study two other cases: G(A; B) = A1=2(A−1=2BA−1=2)1=2A1=2; log A+log B the Pusz-Woronowicz geometric mean, and G(A; B) = exp 2 ; the log Euclidean mean. With these choices d(A; B) is no longer a metric, but it turns out that d2(A; B) is a divergence. We establish some (strict) convexity properties of these divergences. We obtain characterisations of barycentres of m positive definite matrices with respect to these distance measures. 1. Introduction Let p and q be two discrete probability distributions; i.e. p = (p1; : : : ; pn) and q = (q1; : : : ; qn) are n -vectors with nonnegative coordinates such that P P pi = qi = 1: The Hellinger distance between p and q is the Euclidean norm of the difference between the square roots of p and q ; i.e. 1=2 1=2 p p hX p p 2i hX X p i d(p; q) = k p− qk2 = ( pi − qi) = (pi + qi) − 2 piqi : (1) This distance and its continuous version, are much used in statistics, where it is customary to take d (p; q) = p1 d(p; q) as the definition of the Hellinger H 2 distance.
    [Show full text]
  • Numerical Linear Algebra and Matrix Analysis
    Numerical Linear Algebra and Matrix Analysis Higham, Nicholas J. 2015 MIMS EPrint: 2015.103 Manchester Institute for Mathematical Sciences School of Mathematics The University of Manchester Reports available from: http://eprints.maths.manchester.ac.uk/ And by contacting: The MIMS Secretary School of Mathematics The University of Manchester Manchester, M13 9PL, UK ISSN 1749-9097 1 exploitation of matrix structure (such as sparsity, sym- Numerical Linear Algebra and metry, and definiteness), and the design of algorithms y Matrix Analysis to exploit evolving computer architectures. Nicholas J. Higham Throughout the article, uppercase letters are used for matrices and lower case letters for vectors and scalars. Matrices are ubiquitous in applied mathematics. Matrices and vectors are assumed to be complex, unless ∗ Ordinary differential equations (ODEs) and partial dif- otherwise stated, and A = (aji) denotes the conjugate ferential equations (PDEs) are solved numerically by transpose of A = (aij ). An unsubscripted norm k · k finite difference or finite element methods, which lead denotes a general vector norm and the corresponding to systems of linear equations or matrix eigenvalue subordinate matrix norm. Particular norms used here problems. Nonlinear equations and optimization prob- are the 2-norm k · k2 and the Frobenius norm k · kF . lems are typically solved using linear or quadratic The notation “i = 1: n” means that the integer variable models, which again lead to linear systems. i takes on the values 1; 2; : : : ; n. Solving linear systems of equations is an ancient task, undertaken by the Chinese around 1AD, but the study 1 Nonsingularity and Conditioning of matrices per se is relatively recent, originating with Arthur Cayley’s 1858 “A Memoir on the Theory of Matri- Nonsingularity of a matrix is a key requirement in many ces”.
    [Show full text]
  • An Algebraic Approach to Harmonic Polynomials on S3
    AN ALGEBRAIC APPROACH TO HARMONIC POLYNOMIALS ON S3 Kevin Mandira Limanta A thesis in fulfillment of the requirements for the degree of Master of Science (by Research) SCHOOL OF MATHEMATICS AND STATISTICS FACULTY OF SCIENCE UNIVERSITY OF NEW SOUTH WALES June 2017 ORIGINALITY STATEMENT 'I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.' Signed .... Date Show me your ways, Lord, teach me your paths. Guide me in your truth and teach me, for you are God my Savior, and my hope is in you all day long. { Psalm 25:4-5 { i This is dedicated for you, Papa. ii Acknowledgement This thesis is the the result of my two years research in the School of Mathematics and Statistics, University of New South Wales. I learned quite a number of life lessons throughout my entire study here, for which I am very grateful of.
    [Show full text]